python高级-22.pandas-数据聚合和分组计算

import numpy as np
import pandas as pd
from pandas import Series,DataFrame
from numpy import nan as NA
import matplotlib.pyplot as plt
plt.rcParams['font.sans-serif'] = ['Microsoft YaHei']

df1=DataFrame(
    {
   
        'key1':list('aabba'),
        'key2':'one,two,one,two,one'.split(','),
        'data1':np.random.randint(0,10,5),
        'data2':np.random.randint(0,10,5)
    }
)
df1

key1	key2	data1	data2
0	a	one	7	2
1	a	two	4	5
2	b	one	3	7
3	b	two	9	8
4	a	one	1	2
s1=df1['data1']

Series.groupby

按照某个series 进行分组，返回一个分组后的对象

s1.groupby(df1['key1'])

<pandas.core.groupby.generic.SeriesGroupBy object at 0x0000017FF3763F28>

# 可以用list()转换，查看结构，可以知道分组对象的每个值都是分组依据和分组结果
list(s1.groupby(df1.key1))

[('a',
  0    7
  1    4
  4    1
  Name: data1, dtype: int32),
 ('b',
  2    3
  3    9
  Name: data1, dtype: int32)]

我们的分组对象是可以直接使用聚合（统计）函数

s1.groupby(df1.key1).sum()

key1
a    12
b    12
Name: data1, dtype: int32

s1.groupby(df1.key1).mean()

key1
a    4
b    6
Name: data1, dtype: int32

s1.groupby(df1.key1).count()

key1
a    3
b    2
Name: data1, dtype: int64

DataFrame.groupby

df1

key1	key2	data1	data2
0	a	one	7	2
1	a	two	4	5
2	b	one	3	7
3	b	two	9	8
4	a	one	1	2

# 用匹配的值
df1.groupby(df1.key2)

<pandas.core.groupby.generic.DataFrameGroupBy object at 0x0000017FF379FEB8>

list(df1.groupby(df1.key2))

[('one',
    key1 key2  data1  data2
  0    a  one      7      2
  2    b  one      3      7
  4    a  one      1      2),
 ('two',
    key1 key2  data1  data2
  1    a  two      4      5
  3    b  two      9      8)]

# 如果分组依据来自本身的某列，直接用列名称
df1.groupby('key2')

<pandas.core.groupby.generic.DataFrameGroupBy object at 0x0000017FF37B5668>

# 可以用其他独立的Series进行分组
df1.groupby(Series(['c','c','d','d','d']))

<pandas.core.groupby.generic.DataFrameGroupBy object at 0x0000017FF37B5EF0>

list(df1.groupby(Series(['c','c','d','d','d'])))

[('c',
    key1 key2  data1  data2
  0    a  one      7      2
  1    a  two      4      5),
 ('d',
    key1 key2  data1  data2
  2    b  one      3      7
  3    b  two      9      8
  4    a  one      1      2)]

# 可以直接使用聚合函数
df1.groupby(Series(['c','c','d','d','d'])).sum()

data1	data2
c	11	7
d	13	17

df1.groupby(Series(['c','c','d','d','d'])).max()

key1	key2	data1	data2
c	a	two	7	5
d	b	two	9	8

使用多个依据进行分组

DataFrame.groupby([Series1,Series2,…])
如果Series(mapper)来源本身，可以直接使用列名称
DataFrame.groupby([columnName1,columnName2,…])

df1.groupby([df1.key1,df1.key2])

<pandas.